> ## Documentation Index
> Fetch the complete documentation index at: https://cntillion.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Churn reasons

> Find out why customers leave, from every conversation, and group them by reason

When a customer tells your agent they are leaving, the reason is worth more than
the conversation. Churn detection reads each finished conversation, records why
the customer is leaving on their contact, and groups your customers by reason so
you can act on each group.

Open your agent, then **Contacts**, **Churn reasons**.

## Turning it on

Switch on **Detect churn reasons in conversations**. It is off until you do,
because the agent reads every finished conversation once, and each read uses AI
tokens.

A conversation counts as finished once it has ended, or has been quiet for 30
minutes. The agent reads conversations from the website chat, voice calls and
every connected channel, from the last 30 days. A conversation that continues
later is read again. Playground conversations and test calls are never read.

## What it records

For each conversation the agent notes whether the customer said they are
leaving, cancelling or closing their account, or asked how to. A complaint on
its own does not count. When they did, it picks the category that best fits
their reason and keeps a one-line summary in their own words.

The reason is saved on the customer's contact in three attributes, which appear
as columns in Contacts:

| Attribute | Holds |
| - | - |
| Churn reason | The category |
| Churn reason detail | What they said, summarised |
| Churn reason date | When they said it |

A newer reason replaces an older one, never the other way round.

The agent needs to know who the customer is to save the reason on a contact.
Channels such as WhatsApp, SMS and email identify the customer by their number
or address, and a contact is created if none exists. On the website, the visitor
must be [signed in](https://cntillion.com/docs/customer-data/identity-verification.md). A reason from an
anonymous visitor is counted, but belongs to no contact.

If a [CRM is connected](https://cntillion.com/docs/customer-data/crm-integrations.md) with write-back on, the
reason and a summary of the conversation are also logged on the customer's CRM
record.

## Categories

The default categories are:

* Changed job / stopped receiving salary
* Missing feature
* Didn't know a feature existed
* Fees / pricing
* Trust / security
* Support experience
* Other

**Edit categories** renames them, adds your own and removes the ones you do not
need. Each category has a short description, which is what the agent reads when
it decides where a reason belongs. **Other** always stays, for reasons that fit
nothing. **Restore defaults** brings the list above back.

## Customers by reason

The page counts the contacts in each category. Choose one to list its customers
with what they said and when. Each group shows the filter a campaign uses to
target exactly those customers, `churn_reason = <category>`, and **Start a
campaign** opens a new [Outbound campaign](https://cntillion.com/docs/campaigns.md) with that group as its
audience.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.